CEA:基于弱势种群保护抗早熟的聚类淘汰算法

CEA: Algorithm for Cluster- Elutriating based on Minority Protection

作者：向剑平(四川大学计算机学院；遵义师范学院计算机科学系)；唐常杰(四川大学计算机学院)；陈瑜(四川大学计算机学院)；王悦(四川大学计算机学院)；杨宁(四川大学计算机学院)

Author：Xiang jian-ping(School of Computer Science, Sichuan University,)；TANG Chang-Jie(School of Computer Science, Sichuan University)；CHEN Yu(School of Computer Science, Sichuan University)；WANG Yue(School of Computer Science, Sichuan University)；YANG Ning(School of Computer Science, Sichuan University)

收稿日期：2008-12-07 年卷（期）页码：2009,41(5):146-150

期刊名称：工程科学与技术

Journal Name：Advanced Engineering Sciences

关键字：CEA算法；最远临近聚类；β- cluster；个体适应度相似

Key words：cluster- elutriate Algorithm;the farthest neighbor clustering;β- cluster;individual similarity

基金项目：国家自然科学基金

中文摘要

传统基因表达式编程算法(GEP)决定个体遗传权时过分依赖适应度，忽略了个体间相互关系，造成GEP算法易早熟而影响进化效率。为克服该问题，本文：从理论上研究了造成GEP早熟的原因，并根据研究结果提出弱势种群保护抗早熟的聚类淘汰算法CEA（cluster- elutriate Algorithm）；定义β- cluster及相关概念；用种群所含不同簇的数量来度量种群的多样性达到保护弱势种群。利用概率手段详细分析了个体参与下一代的机率。实验表明，基于CEＡ的算法能很好的防止GEP函数发现时的早熟现象，且极大的提高了函数发现效率。

英文摘要

Abstract: In traditional Gene Expression Programming (GEP), the survival of individuals depends on their fitness, and the relationships between them are ignored. This may affect the evolution efficiency. To tackle these problems, this paper analyzes the cause of premature in GEP, introduces a Cluster-Elutriate Algorithm (CEA) based on minority-protection strategy, and develops a group of concepts, such as, β- cluster. The results of experience shows that the efficiency of GEP is improved based on GEA.

【关闭】

论文摘要

CEA:基于弱势种群保护抗早熟的聚类淘汰算法

CEA: Algorithm for Cluster- Elutriating based on Minority Protection